In this study we tested the longitudinal associations of PA with other health behaviours, namely alcohol consumption, sleep quality, nutrition quality, diet quantity, and sedentary time during the COVID-19 pandemic. Increases in PA were associated with increases in alcohol consumption, sleep quality, nutrition quality, and time spent doing sedentary activities. However, changes in PA were not associated with changes in diet quantity. These associations were independent of all time-invariant confounders as well as a range of time-varying confounders and other health behaviours. There was limited evidence that the longitudinal association between PA and specific health behaviours differed according to baseline characteristics. Only participants who reported being overweight at the start of lockdown (according to their BMI) had significant increases in diet quantity in tandem with increases in PA, unlike participants who were not overweight.
Our sample showed a modest improvement in nutrition quality in parallel with increases in PA, although there were no associations between PA and diet quantity overall. Most studies have reported near-ubiquitous increases in snacking and overall food consumption during lockdowns [56–58]. This population-level increase in food consumption may have masked any possible associations between PA and diet quantity observed in this study. Nonetheless, the association of PA with improved nutrition is in line with pre-pandemic studies comparing the diets of active and inactive individuals, which report differences in food preferences; active individuals favour lower fat savoury foods despite having increased food demands overall [7, 8]. Mechanistically, acute exercise has been linked to reductions in levels of hormones controlling appetite (e.g., ghrelin) and may contribute to diet and nutrition changes, especially in response to stressful situations such as a lockdown [59, 60]. Worse food choices including snacking can also result from stress-induced cortisol spikes, and this spiking can be attenuated by acute bouts of PA [60]. Alongside existing evidence, our findings suggest that PA may have continued to help to support healthy dietary behaviours during the stresses of lockdown.
The relationship between increased PA and enhanced sleep quality observed in this study is likely bidirectional. Both single acute and regular bouts of exercise have been observed to improve individuals’ sleep by acting on a combination of circadian, metabolic, thermoregulatory, mood and endocrine pathways [11]. This is a positive feedback relationship, as good quality sleep supports both mood and daily energy levels which can indirectly improve engagement in PA [11]. It has previously been reported that altered work patterns during the COVID-19 pandemic led to widespread negative changes in sleep timings and increases in daytime napping for many, which may have longer-term negative consequences for psychological health [61]. Our findings suggest that encouraging PA could support sleep quality and counter this negative impact of the pandemic.
The relationship between increased PA and increased alcohol consumption in our sample may in part be attributable to the substantial overall increase in alcohol use in the population during the COVID-19 pandemic, coupled with changes in recreational time available [43, 49, 62, 63]. For many individuals, the increased leisure time available during the COVID-19 lockdown may have led to boredom, which may in turn have increased both PA and alcohol consumption. A recent study examining the motivations of adolescents to increase their PA during lockdown identified ‘increased time’ and ‘boredom’ as substantial factors [64]. Boredom and the feeling of the slower passage of time have also been reported in at least one cohort during the pandemic [65]. Additionally, boredom has been associated with greater binge drinking, as well as both negative and positive changes to participant PA during the pandemic [48]. However, for others, increased free time may have been an opportunity and motivation to make positive lifestyle changes [66]. Differences in individuals’ backgrounds and circumstances likely influenced their health behaviours during lockdown. For instance, individuals with more available time due to working remotely or being on furlough may have chosen to spend their additional time in both positive and negative ways such as exercising, cooking more healthily, getting more sleep, but also consuming more alcohol and engaging in more sedentary activities, as seen in this study.
The observed association between increased PA and increased sedentary time is not surprising given the unique context of this study. In the enforced lockdown, sedentary hours may have replaced other low intensity PA such as walking to, from, and within work, taking children to school, or other outdoor chores and errands [17, 27, 31]. This increase in overall sedentary time is likely to have occurred for most people, even if they also did more dedicated PA (e.g., going on a walk or run, or working out at home) which is measured in this study. Additionally, increases in PA for exercise are often followed by increases in sedentary time in the following days due to acute fatigue and recovery [67]. This is most clear in exercise interventions for weight loss which often report a compensation phenomenon, either in participant’s diets or sedentary time, which partially offsets the benefits of increased PA [68]. This could be particularly relevant to individuals who used lockdown to improve their PA engagement. For example, participants in this study who reported being overweight had a greater increase in diet quantity with increased PA than participants who were not overweight. However, it is worth noting that the activities considered sedentary in this study are not intrinsically negative. Distinctions between positive and negative sedentary time have been drawn in previous literature which found differences in the associations of health outcomes with TV-viewing versus motorised commuting or sitting for work. Indeed, other evidence from this cohort has highlighted that some sedentary behaviours, such as remotely engaging in the creative arts, provided a way of coping with the stress of lockdowns [20, 69] and were associated with improvements in depression, anxiety and life satisfaction [70]. Our sensitivity analyses may support this finding, given that increases in PA were associated with increases in time spent on these creative arts activities. Nonetheless all types of sedentary time, even those that were screen-based and often considered negative, remained positively associated with PA.
This study has several strengths, including the use of a fixed effects approach and a large sample size with weekly measures across a period of 22 consecutive weeks during the first UK national lockdown and following easing of restrictions. However, it is important to acknowledge that findings from fixed effects models have limited generalisability to the population [71, 72], and the results of this study should be prefaced with 'within those whom change'. Like other studies during the COVID-19 pandemic, there is a lack of pre-pandemic measures, limiting inferences regarding the magnitude of the associations observed in this study before versus during the pandemic. Further limitations include the use of a self-selected sample. Despite this, the large sample shows wide heterogeneity and good stratification across all major socio-demographic groups. Further efforts to improve sample representativeness included the use of weighting to align with national population statistics, making the sample comparable to the sample of another nationally representative study [73]. However, despite all efforts to make our sample inclusive and representative of the adult population, we cannot rule out the possibility of potential biases due to omitting other demographic factors that could be associated with survey participation in the weighting process. Additionally, our survey did not utilise a validated tool of daily PA, instead focussing on ‘dedicated’ or ‘leisure-time’ PA. Although this approach is widely used, it is less reflective of total daily movement. Lastly, we cannot rule out the possibility of reporting biases including social desirability in participant’s responses given the reliance on self-report tools, although use of a self-completed online survey may in-part have buffered this risk.